--hide-reasoning and hide_reasoning=True parameters

[simonw]

Previously we had a -R/--no-reasoning option to llm prompt which was misleadingly named - it only suppressed output of reasoning text, it did not turn reasoning off - and was not passed through to the models.

As of #1435 the default OpenAI model plugin requests "summary": "auto" reasoning summaries. You can hide these with -R but they are still requested.

This change passes hide_reasoning=True through on the Prompt object, such that plugins can opt not to request visible reasoning tokens.

[simonw]

• Docs for plugin implementors: https://github.com/simonw/llm/blob/e020d7c8c18425200addd5b91af76c176691051f/docs/plugins/advanced-model-plugins.md#reasoning-tokens
• End-user Python API docs: https://github.com/simonw/llm/blob/2acc0aa64974316e491d5008a6d1f360ddc2f98d/docs/python-api.md#hiding-reasoning-output